Lecture #25 ggplot III

ggplot (continued)

Preliminaries

library(ggplot2)
library(ggthemes)
library(patchwork)
library(colorblindr)
library(cowplot)
library(colorspace)
library(wesanderson)
library(ggsci)
library(TeachingDemos)

char2seed("Dark Star")
d <- mpg

Bar plots

# use to plot the counts of rows for a categorical variable
table(d$drv)

## 
##   4   f   r 
## 103 106  25

p1 <- ggplot(d,aes(x=drv)) + geom_bar(color="black",fill="cornsilk")
print(p1)

# aesthetic mapping gives multiple groups for each bar
p1 <- ggplot(d,aes(x=drv,fill=fl)) + geom_bar()
print(p1)

# stacked, but need to adjust color transparency, which is "alpha"
p1 <- ggplot(d,aes(x=drv,fill=fl)) + geom_bar(alpha = 0.3, position="identity")
print(p1)

# better to use position = fill for stacking, but with equivalent height
p1 <- ggplot(d,aes(x=drv,fill=fl)) + geom_bar(position="fill")
print(p1)

# best to use position = dodge for multiple bars
p1 <- ggplot(d,aes(x=drv,fill=fl)) + geom_bar(position="dodge",color="black",size=1)
print(p1)

# more typical "bar plot" has heights as the values themselves
dTiny <- tapply(X=d$hwy,INDEX=as.factor(d$fl),FUN=mean) #calculate the means
dTiny <- data.frame(hwy=dTiny) # create a single-column data frame
dTiny <- cbind(fl=row.names(dTiny),dTiny) # 

p2 <- ggplot(dTiny, aes(x=fl,y=hwy,fill=fl)) +
  geom_col()
print(p2)

#### Use a box plot instead of standard "means" bars!

# basic boxplot is simple and informative
p1 <- ggplot(d,aes(x=fl,y=hwy,fill=fl)) +
               geom_boxplot()
print(p1)

# now overlay the raw data
p1 <- ggplot(d,aes(x=fl,y=hwy)) +
               geom_boxplot(fill="thistle",outlier.shape=NA) + 
  # geom_point()
  geom_point(position=position_jitter(width=0.1,height=0.7),color="grey60",size=2)

print(p1)

### Using Colors

1. Aesthetics

1. Colors that are attractive - large geoms (fills) - pale colors - small geoms(lines,points) - bright colors
2. Colors that highlight elements - pale, grey to de-emphasize - bright or saturated colors to emphasize
3. Colors that are visible to the color blind
4. Colors that convert well to black and white

2. Information content

1. Discrete scale

• colors to group similar treatments
• neutral colors (black,grey,white) to indicate control groups
• Symbolic colors (heat=red, cool = blue, photosynthesis/growth=green, oligotrophic=blue, eutrophic=brown, infected=red)
• Colors that map to chemical stains or gels, or even colors of organisms

2. Continuous scale

• monochromatic (differing shades of 1 color)
• 2 tone chromatic scale (from color x to color y)
• 3 tone divergent scale (from color x through color y to color z)

3. Use color information within and between graphs

3. show color names, hex in base R
4. show color schemes in colorbrewer

Color visualizations

my_cols <- c("thistle","tomato","cornsilk","cyan","chocolate")
demoplot(my_cols,"map")

demoplot(my_cols,"bar")

demoplot(my_cols,"scatter")

demoplot(my_cols,"heatmap")

demoplot(my_cols,"spine")

demoplot(my_cols,"perspective")

Color customizations

d <- mpg
# --------- discrete classification
# scale_fill_manual for boxplots,bars
# scale_color_manual for points, lines

# boxplot no color
p_fil <- ggplot(d,aes(x=as.factor(cyl),y=cty))
p_fil + geom_boxplot()

# boxplot default ggplot fill
p_fil <- ggplot(d,aes(x=as.factor(cyl),y=cty,fill=as.factor(cyl))) + geom_boxplot()
plot(p_fil)

# create custom color palette
my_cols <- c("red","brown","blue","orange")


# boxplot with custom colors for fill
p_fil + scale_fill_manual(values=my_cols)

# scatterplot with no color
p_col <- ggplot(d,aes(x=displ,y=cty))
p_col + geom_point(size=3)

# scatterplot default ggplot colors
p_col <- ggplot(d,aes(x=displ,y=cty,col=as.factor(cyl))) + geom_point(size=3)
plot(p_col)

# scatterplot with custom colors for point color
p_col + scale_color_manual(values=my_cols)

# continuous classification (color gradient)

# default color gradient
p_grad <- ggplot(d,aes(x=displ,y=cty,col=hwy)) + geom_point(size=3)
plot(p_grad)

# custom sequential gradient (2-colors)
p_grad + scale_color_gradient(low="green", high="red")

# custom diverging gradient (3-colors)
mid <- median(d$cty)
p_grad + scale_color_gradient2(midpoint=mid,
                               low="blue",
                               mid="white",
                               high="red")

Working with color and grey

d <- mpg # use built in mpg data frame
p1 <- ggplot(d,aes(x=fl,y=hwy,group=fl)) 
p1 + geom_boxplot() # no color

p1 + geom_boxplot(fill="red") # single color (no info)

# create a palette
myColors <- c("red","green","pink","blue","orange") #no info
p1 + geom_boxplot(fill=myColors)

myColors <- c("red","green","green","blue","orange") # info
p1 + geom_boxplot(fill=myColors)

p1 + geom_boxplot(fill=gray(seq(from=0.1,to=0.9,length=5)))

# using alpha transparency for histograms
x1 <- rnorm(n=100,mean=0)
x2 <- rnorm(n=100,mean=3)
dFrame <- data.frame(v1=c(x1,x2))
lab <- rep(c("Control","Treatment"),each=100)
dFrame <- cbind(dFrame,lab)
str(dFrame)

## 'data.frame':    200 obs. of  2 variables:
##  $ v1 : num  -0.0306 1.4022 0.166 2.5987 -1.1781 ...
##  $ lab: Factor w/ 2 levels "Control","Treatment": 1 1 1 1 1 1 1 1 1 1 ...

h1 <- ggplot(dFrame,aes(x=v1,fill=lab))
h1 + geom_histogram(position="identity",alpha=0.5,color="black") 

# now go out to web and show them the palette page
p1 + geom_boxplot(fill=wes_palettes[["Royal2"]])

p1 + geom_boxplot(fill=c(grey(0.5),canva_palettes[[1]]))

# use scale fill manual

p2 <- ggplot(d,aes(x=fl,y=hwy,fill=fl)) + geom_boxplot()+ scale_fill_manual(values=wes_palettes[["Darjeeling1"]])
print(p2)

myCanvas <- canva_palettes[["Fresh and bright"]]
myCanvas <- c(gray(0.6),myCanvas)
p2 + geom_boxplot() + scale_fill_manual(values=myCanvas)

Colorbrewer page

# visit colorbrewer page for demo
p2 + geom_boxplot() + scale_fill_brewer(palette="Blues")

## Scale for 'fill' is already present. Adding
## another scale for 'fill', which will replace
## the existing scale.

# slightly different scheme for points, which have color

p3 <- ggplot(d,aes(x=displ,y=hwy,color=fl)) +
  geom_point()   +
   scale_color_brewer(palette="Spectral")
print(p3)

Continuous variables

p3 <- ggplot(d,aes(x=displ,y=hwy,color=cty)) +
  geom_point()    
print(p3)

# use scale_color_gradient to change low and high colors
  # scale_color_gradient(
p3 +
  scale_color_gradient(low="red",high="blue")

# use scale_color_gradient2 for a 3 color gradient
z=mean(d$cty)
p3 +
  scale_color_gradient2(midpoint=z,low="red",mid="seagreen",high="cyan",space="Lab")

# use scale_color_gradientn for multicolored changes
p3+scale_color_gradientn(colours = rainbow(5))

p3+scale_color_gradientn(colours = myCanvas)

Making a heat map

xVar <- 1:30
yVar <- 1:5
myData <- expand.grid(xVar=xVar,yVar=yVar)
head(myData)

##   xVar yVar
## 1    1    1
## 2    2    1
## 3    3    1
## 4    4    1
## 5    5    1
## 6    6    1

zVar <- myData$xVar + myData$yVar + 2*rnorm(n=150)
myData <- cbind(myData,zVar)
head(myData)

##   xVar yVar zVar
## 1    1    1 1.93
## 2    2    1 4.78
## 3    3    1 4.30
## 4    4    1 3.63
## 5    5    1 6.51
## 6    6    1 5.77

p4 <- ggplot(myData,aes(x=xVar,y=yVar,fill=zVar))
p4 + geom_tile()

p4 <- ggplot(myData,aes(x=xVar,y=yVar,fill=zVar))
p4 + geom_tile() +
scale_fill_gradient2(midpoint=19,low="brown",mid=grey(0.8),high="darkblue")

#### New viridis color scale

p4 + geom_tile() + scale_fill_viridis_c()

# options viridis, cividis, magma, inferno, plasma
p4 + geom_tile() + scale_fill_viridis_c(option="inferno")

p4 <- p4 + geom_tile() + scale_fill_viridis_c() 
p4des<-edit_colors(p4, desaturate)
ggdraw(p4des)